JP4643993B2 - Limit order execution probability calculation system and execution probability calculation program - Google Patents

Description

  The present invention relates to a limit order execution probability calculation system and execution probability calculation program, and in particular, when a limit order of a specific price / quantity related to a specific issue is issued to the stock market, it is possible to execute with a certain probability. The present invention relates to a technique for automatically calculating whether or not there is.

Traders in institutional investors, etc., make various orders every day to the stock market, and are struggling to secure margins through stock trading.
Stock orders include a market order that specifies only the brand and quantity and ignores the price, and a limit order that specifies the transaction price along with the brand and quantity. In some cases, it is mainly performed, and in order to secure profits by trading, it is necessary to make sure that trading is achieved at a targeted price. It can be said.
For this reason, the trader is required to determine the price at which the transaction will be finalized and place an order, but if it can recognize the execution probability to some extent in advance, compared to relying solely on personal intuition and experience, More objective and rational investment activities are possible.
Also, in order to realize that a part of stock trading transactions are automatically processed by a computer, it is necessary to establish a model for calculating the execution probability of a specific limit order as a precondition.

Hitoshi Sakurai and Yukihiro Asano “The Latest Theory of Asset Management” (Nihon Keizai Shimbun, published in 2002) Chapter 5 “Execution Probability of Limit Orders” (P127-155)

  By the way, trading transactions in the stock market are based on a changing environment and a combination of various conditions. Therefore, in order to build a model for calculating the execution probability of limit orders, the limit in the market is assumed as a premise. It is necessary to collect a large number of filled and unfilled orders, and statistically analyze the contents of each order, the environment and conditions at the time of placing the order, and the correlation between the results (success or failure).

However, since detailed information on individual transactions on the stock exchange is not disclosed to the public, basic data for constructing a reliable model is absolutely lacking.
Of course, if you are an institutional investor or a securities company, you have accumulated a certain amount of detailed data on the transactions that you deal with, but even if you are an institutional investor or a securities company, In general, there is a bias in the ordering pattern, and there is not enough data to explore market trends and mechanisms.

For this reason, there has been little awareness of the problem of calculating the execution probability of limit orders.
The only non-patent document 1 described above reproduces the ordering behavior of market participants from the tick data of the Tokyo Stock Exchange, adds some consideration to the determinants of execution probability, and based on these, the execution probability of limit orders However, it does not deviate from the scope of academic scholarship and does not have enough concreteness and completeness to be applied immediately to real transactions.

  The present invention has been devised in view of the above-mentioned present situation related to calculation of execution probability of limit orders, and generates a probability model reflecting actual market trends with high accuracy, and based on this, a specific limit order is generated. The purpose is to provide a technology that can specifically calculate the execution probability.

In order to achieve the above object, the system for calculating the execution probability of a limit order according to claim 1 includes at least stock information including a stock name, date, price, quantity transmitted from a stock exchange computer, and at least stock information. Means to sequentially store quotation information including brand name, date, trading type, price, quantity in the market data storage unit, and extract limit order information based on the time series change of the quotation information, at least the brand, date Limit order extraction means for associating the trade type, price, quantity with a unique ID and storing it in the limit order storage unit, and extracting cancellation information of limit order based on the time series change of the above execution information and quotation information, at least Cancel extraction means for associating the brand, date and time, trade type, price and quantity with a unique ID and storing them in the cancel storage unit, and the above contract information and cancel information. Based on the information, execution, cancellation, uncommitted determination is made for each limit order, each determination result is stored in the determination result storage unit, and predetermined attribute information relating to the execution limit and uncommitted limit orders Means for calculating based on the determination result and information stored in the market data storage unit, means for calculating the remaining time of the transaction based on the date and time of each limit order and the transaction end time of the stock exchange, and Means for generating a probability model for each brand by statistically processing the determination result of the promised or unfilled order related to the limit order, the quantity, the remaining time of the transaction, and the attribute information, and storing the probability model in the probability model storage unit When a limit order specifying a brand, trade type, price, and quantity is input, a means for selecting a probability model to be applied based on the brand, and a current required order of the limit order Means for calculating the sex information based on the information stored in the market data storage unit, means for calculating the remaining time of the transaction based on the current time and the transaction end time of the stock exchange, and the brand of the limit order A system for calculating execution probability by applying the transaction type, price, quantity, remaining time of transaction, and necessary attribute information to the probability model, wherein the limit order extraction means is: When the quantity related to the quotation information of a specific brand, trade type, and price at a certain time t stored in the market data storage unit increases at a later time t + 1, the brand, date / time, trade type, price Then, the limit order information for the increased quantity is extracted, and the cancel extraction means is configured to determine the quantity related to the specific brand, trade type, price quotation information at a certain time t stored in the market data storage unit. It decreases at time t + 1 after, and if not present contract information decreases minute quantities in the meantime, the stock, date, sale type, price, extracts cancel information reduction minute quantities, the contract judgment The means determines the success or failure of the execution by allocating a constant amount to each limit order in order of priority in terms of time and price, and assigns a cancellation quantity in order of subordination in terms of time and price. It is characterized in that success / failure is determined, and a limit order that has no determination of execution or cancellation is determined as unsuccessful.

A probability model may be generated for each brand and trading type and stored in the probability model storage unit, and a probability model to be applied may be selected according to the input limit brand and trading type. .
In this way, by preparing a probability model optimized for each trading type as well as the brand, switching the probability model to be applied according to the trading type of the limit order when calculating the execution probability, the more accurate execution probability It becomes possible to guide.

  The limit order execution probability calculation system described in claim 2 is characterized in that at least one of the attribute information is a cumulative balance of orders that are prioritized in terms of time and price with respect to the target limit order.

  In the limit order execution probability calculation system according to claim 3, at least one of the above attribute information is a median divergence ratio obtained by dividing a price difference between a price of a limit order and a median price of a target by a price of a limit order. It is characterized by that.

The limit order execution probability calculation program according to claim 4, wherein the computer executes a computer, execution information including at least stock name, date / time, price, and quantity transmitted from a stock exchange computer, and at least stock name, date / time, Means for sequentially storing quotation information including trading type, price, quantity in the market data storage unit, extracting limit order information based on the time series change of the above quotation information, at least its brand, date, trading type, price, Limit order extraction means for associating the quantity with a unique ID and storing it in the limit order storage unit, extracting the cancel information of limit order based on the time series change of the above execution information and quotation information, at least its brand, date, trading type , Cancel extraction means for associating a price and quantity with a unique ID and storing them in a cancellation storage unit, based on the above execution information and cancellation information, Executes determination, cancellation, and uncommitted determination regarding the limit order, and stores each determination result in the determination result storage unit, and predetermined attribute information related to the limit order for execution and uncommitted, the determination result and the above Means for calculating based on information stored in the market data storage unit, means for calculating the remaining time of trading based on the date and time of each limit order and the trading end time of the stock exchange, execution or uncommitted for each limit order The determination result, the quantity, the remaining time of the transaction, and the above attribute information are statistically processed to generate a probability model for each brand, a means for storing this in the probability model storage unit, brand, trading type, price, When a limit order specifying a quantity is input, a means for selecting a probability model to be applied based on the stock, and attribute information required at the present time of the limit order Means to calculate based on information stored in the storage unit, means to calculate the remaining time of the transaction based on the current time and the transaction end time of the stock exchange, brand of the limit order, trading type, price, quantity, transaction The remaining time and necessary attribute information are applied to the probability model to function as means for calculating execution probability, and the limit order extraction means is stored in the market data storage unit. When the quantity related to the quotation information of a specific brand, trade type and price at time t increases at time t + 1 later, limit order information for the brand, date and time, trade type, price, and increased quantity is extracted. Then, the cancel extraction means is configured such that the quantity related to the specific brand, trade type and price quotation information stored at the time t stored in the market data storage unit is later at time t + 1. If the contract information for the reduced quantity does not exist during this period, the stock, date / time, trade type, price, and cancellation information for the reduced quantity are extracted. In addition to determining the success or failure of the execution by assigning contracted amounts in order of priority in terms of time and price, the success or failure of the cancellation is determined by assigning cancellation quantities in the order of subordination in terms of time and price. Further, it is characterized in that a limit order without any determination of cancellation and cancellation is determined as unfilled.

The limit order execution probability calculation system according to claim 1 and the execution probability calculation program according to claim 4, wherein the limit order is determined from a time series change of fragmentary quotation information provided from a computer of a stock exchange. The information is virtually extracted, and the cancellation information is virtually extracted from the time series change of the contract information and the quotation information, and the cancellation order and the success or failure of the execution are determined for each limit order. A probability model for calculating execution probability is generated by statistically analyzing the attribute information related to, and a mechanism for calculating the execution probability of the actual limit order based on the probability model is provided.
Therefore, even if the individual limit order information itself extracted is not the same as the actual transaction, it comprehensively embodies the movement in the market at that time.
In addition, by broadly extracting virtual limit order information from the transaction information provided by the stock exchange, it becomes possible to collect a huge number of contract cases and uncommitted cases, and the attribute information A probability model that strongly reflects the trend of the trading market can be generated by appropriately selecting factors that influence the result of the promise / uncommitted limit order.
As a result, the execution probability can be calculated with high accuracy by inputting the current data and the actual limit order data into the probability model.

  The limit order execution probability calculation system according to claim 2 uses a so-called cumulative priority book balance that has a great influence on the limit order execution probability as one of the basic data for generating the probability model, and at the time of execution probability calculation, Since the accumulated priority book balance in is reflected in the calculation result, it is possible to derive an execution probability that is more appropriate to the actual transaction status.

  The limit order execution probability calculation system according to claim 3 uses the divergence rate between the order price and the middle price, which has a great influence on the limit order execution probability, as one of the basic data for generating the probability model, At the time of calculating the probability, the current mid-value divergence rate is reflected in the calculation result, so that it is possible to derive an execution probability that is more appropriate to the actual transaction status.

  FIG. 1 is a block diagram showing the functional configuration of a limit order execution probability calculation system 10 according to the present invention. The market data acquisition unit 12, the market data storage unit 14, the limit order extraction unit 16, and the limit order DB 18 A cancellation extraction unit 20, a cancellation DB 22, a contract determination unit 24, a determination result DB 26, an order attribute calculation unit 28, an order attribute DB 30, a probability model generation unit 32, a probability model DB 34, and an execution probability calculation. Part 36.

The market data acquisition unit 12, limit order extraction unit 16, cancellation extraction unit 20, execution determination unit 24, order attribute calculation unit 28, probability model generation unit 32, execution probability calculation unit 36, computer CPU, OS and dedicated This is realized by executing necessary processing in accordance with the application program.
The market data storage unit 14, limit order DB 18, cancellation DB 22, determination result DB 26, order attribute DB 30, and probability model DB 34 are provided in the hard disk of the computer.

A market exchange computer 40 is connected to the market data acquisition unit 12 via a communication network 38 such as ISDN, and market transaction data (tick data) that changes from time to time is received from the computer 40 to the market data acquisition unit. Sent to 12.
The execution probability calculating unit 36 is connected to a transaction terminal 42 operated by a trader via a LAN, the Internet, or the like.
The transaction terminal 42 is also connected to the stock exchange computer 40 via the communication network 38, and the transmitted transaction data is processed into a predetermined format and displayed on the display.

FIG. 2 shows the structure of the transaction data transmitted from the stock exchange computer 40. The stock brand, the transaction date and time, the type of information (contract / buy / offer), price and quantity are shown. The items shown are included.
As described above, since the minimum necessary data indicating the result of the transaction is transmitted from the computer 40 in small pieces, each transaction data is aggregated and a screen in which these are shaped to be easily understood by humans is generated. A dedicated application program having a function of displaying this on the display is set up in the transaction terminal 42 in advance.

FIG. 3A shows an example of this screen. A so-called “board” in which the sales price and the sales price regarding the specific brand are two-dimensionally arranged is drawn on the screen.
In this board 44, a price display band 46 is provided between the sales price and the sales price, and the price (indication price) of each price is displayed. Further, the quantity of each sign is expressed by the length and numerical value of the bar graph (the numerical value is omitted for convenience of illustration).
For example, there is a quotation between 102 yen and 106 yen on the ASK side, while there is a quotation between 98 yen and 94 yen on the BID side. FIG. 3A shows that this is the case.
In addition, the best selling price with the lowest price and easy to establish a transaction is referred to as the best selling price, and the best selling price with the highest price and the easily established transaction is referred to as the best selling price.
Further, the price difference between the best bid price and the best bid price is referred to as spread, and the price located between them is referred to as the middle price.
In the case of FIG. 3A, the best bid price is 102 yen, the best bid price is 98 yen, the spread is 4 yen, and the median price is 100 yen.
A plurality of quotation data is generally transmitted from the exchange computer 40. For example, the Tokyo Stock Exchange's computer distributes the top five quotation data (so-called five values) including the best quotation for each sale and purchase.

When the contract data (1) (102 yen / 50 shares) and (2) contract data (98 yen / 30 shares) are newly transmitted to the board 44 in this state, for example, the application program of the transaction terminal 42 Generates the screen of FIG. 3 (b) reflecting these and displays them on the display.
In this new screen board 44, the selling price of 102 yen, which was 120 shares according to the contract data in (1), decreased to 70 shares (-50 shares), and 70 shares in accordance with the contract data in (2). The bid price of 98 yen has decreased to 40 shares (-30 shares).

Next, the processing procedure of the system 10 will be described with reference to the flowchart of FIG.
First, stock trading data that changes from moment to moment is transmitted from the computer 40 to the system 10 while the stock exchange is open.
The market data acquisition unit 12 that has captured the data is processed into a predetermined format and then stored in the market data storage unit 14 (S10).
In the market data storage unit 14, as shown in FIG. 5, a contract information table, a buy price information table, and a sell price information table are provided.
Among these, the transaction data transmitted from the computer 40 is registered in the contract information table as it is, whereas the transaction information transmitted from the computer 40 is stored in the bid information table and the sales information table. Data is registered in a state of being arranged in accordance with the ranking of signs (1 to 5).

In this system 10, an execution probability model for each brand is generated and updated at a predetermined timing (for example, every day at night) using accumulated data for a predetermined period (for example, six months).
First, the limit order extraction unit 16 and the cancellation extraction unit 20 are activated simultaneously with the arrival of the above timing, and the limit order information and the cancellation information are extracted from the accumulated data in the market data storage unit 14, and the limit order DB 18 and the Store in the cancellation DB 22 (S12, S14).

That is, as shown in FIG. 2, the stock exchange computer 40 only transmits a piece of data indicating the result of the transaction.
For example, in the case of transaction data (buy information) in (2), it only indicates that the buy price of 94 yen related to the issue: ABCD has become 100 shares, whether it is a limit order or a market order. Whether it is a result of cancellation or not is not specified.
In the case of (3) transaction data (buy information), it only indicates that the sell price of 103 yen related to the stock: ABCD has reached 50 shares, whether it is a limit order or a market order. Or whether it is the result of a cancellation.
In addition, in the case of the transaction data (contract information) in (1), it only indicates that 50 shares were sold for 103 yen for the issue: ABCD. It is not specified.
For this reason, the system 10 employs a method of preferentially extracting the limit order and its cancellation from the change in the market data as described above based on a predetermined logic.

For example, as shown in FIG. 6, when the bid price of 99 yen for a certain brand increases between time t and t + 1, the limit order extraction unit 16 recognizes the increased stock number (1) as a limit order for buying, As shown in FIG. 7, the brand, date, time, trade type, quote price, and quantity are stored in the limit order DB 18 in association with a specific limit ID.
Also, if a 100 yen bid that did not exist at time t appears at time t + 1, the limit order extraction unit 16 recognizes the increased number of shares (2) as a limit order for selling, and the limit price is the same as above. Store in the order DB 18.

On the other hand, if the selling price of 102 yen is decreasing from the time t to the time t + 1, the cancel extraction unit 20 recognizes the reduced stock number (3) as the sales canceling amount, and as shown in FIG. The brand, date, time, trade type, quotation, and quantity are stored in the cancellation DB 22 in association with a specific cancellation ID.
In addition, even when the bid price of 98 yen that existed at the time t disappears at the time t + 1, the cancel extraction unit 20 recognizes the reduced stock number (4) as a cancellation amount of the purchase, and the cancel DB 22 in the same manner as described above. To store.
However, as shown in FIG. 9, transaction data (contract information) of “contract / 101 yen / 100 shares” exists between time t and time t + 1, and the number of shares related to the quotation of 101 yen at time t + 1. If the decrease has occurred, the cancellation extraction unit 20 recognizes that “the decrease is due to the execution of a buy market order, not the decrease due to the cancellation of the limit order”, and is registered as a cancellation amount in the cancellation DB 22 Make up for what to do.

As described above, after completing the extraction of the limit order and the cancellation of the limit order from the accumulated data for a predetermined period in the market data storage unit 14, the execution of the contract determination unit 24 is started, and the limit order DB 18 For each limit order, it is determined whether or not the contract is successful and cancelled (S16).
That is, neither the contract information stored in the contract information table of the market data storage unit 14 nor the cancel information stored in the cancel DB 22 has a one-to-one correspondence with the limit order registered in the limit order DB 18.
For this reason, there is a problem as to which limit order a certain number of contracted shares or the number of canceled shares should be distributed.
Therefore, the system 10 employs a method of distributing the number of contracted shares and the number of canceled shares to individual limit orders by applying a predetermined distribution rule.

FIG. 10 shows the rules for distributing the commitments.
First, if there is a priority book of 5 shares (a limit order that is superior in time) before the limit order (3 shares) at the time of (1), the execution of 3 shares will be executed at the time of (2). If there is, the priority book is allotted to the priority book, so that the priority book is reduced to two shares.
Next, when a contract of 4 shares occurs at the time of (3), the priority book disappears, and at the same time, 2 shares are allocated to the target limit order, and 1 share remains.
At the time of (4), there are 4 subordinate books that are inferior to the target limit order.
When two shares are executed at the time of (5), all the target limit orders are executed, and at this time, the execution of the limit order is recognized as being completed.

FIG. 11 shows a rule for assigning a cancel amount.
First, if there is a priority book of 5 shares before the limit order (3 shares) at the time of (1) and a subordinate book of 4 shares after that, the cancellation of 3 shares at the time of (2) When there is, the cancellation is allocated to the subordinate book first, and the subordinate book is reduced to one share.
Next, if 2 subordinate books are added at the time of (3) and 5 shares are canceled at the time of (4), the subordinate book will be zero and at the same time the target limit order will be 2 Cancellation of shares is allotted and the remaining 1 share.
When the cancellation of two shares occurs at the time of (5), all the target limit orders are canceled, and at this point, it is recognized that the limit order has been canceled.

In actual transactions, some of the priority books may be canceled, but it is impossible to read them from the limited transaction data provided by the stock exchange computer 40. .
For this reason, in this system 10, in accordance with an empirical rule that “subordinate books with a lower probability of execution have a higher cancellation rate than priority books with a higher probability of execution” The “first out” rule is adopted.

The execution determination unit 24 determines the execution and cancellation of all limit orders according to the above rules, and for those that do not have any determination, it is “Uncommitted”, that is, without executing or canceling the transaction. Qualify as an order left after the end time arrives.
The contract determination unit 24 associates the determination result with each limit ID and stores it in the determination result DB 26 (S18).
FIG. 12 shows an example of the data item. In the determination result item, one of promise, uncommitted, and cancel is recorded.

Next, the order attribute calculation unit 28 is activated to extract a pending or unfilled limit order other than the limit order for which cancellation has been recorded in the determination result DB 26 (S20), and predetermined attribute information relating to each limit order Is calculated based on information in the market data storage unit 14 (S22).
FIG. 13 is a block diagram illustrating the arithmetic processing executed by the order attribute calculation unit 28. Here, the buy best bid quantity calculation, sell best bid quantity calculation, spread calculation, remaining time calculation, cumulative priority book balance calculation, Each process of calculating the median divergence rate, calculating the relative volume at the time of order, calculating the return at the time of order, calculating the relative volume for the latest 5 minutes, and calculating the return for the latest 5 minutes is illustrated.

With regard to “buy best price” and “sell best price”, the order attribute calculation unit 28 extracts the best buy price and the best sell price on the board when each limit order is generated. In general, the probability of execution of a sell limit order will increase while the probability of execution of a sell limit order will increase while the probability of execution of a sell limit order will increase, whereas the probability of execution of a sell limit order will increase. On the other hand, since there is a tendency for the execution probability of the limit order of purchase to increase, these pieces of attribute information are extracted as basic data for calculating the execution probability.
“Spread” is the result of dividing the price difference between the best buy price and the best sell price on the above board by the price of the limit order, and is expressed as a percentage. It is considered that the probability of execution decreases as the spread increases, because the possibility of new limit orders intervening there increases.
“Remaining time” means the remaining time of a transaction when a limit order occurs. Generally, the longer the remaining time, the higher the probability of execution. This remaining time is calculated based on the date and time of each limit order and the transaction end time (advance or close) of the stock exchange.
“Cumulative priority book balance” means the total number of limit orders that are superior in time and price to the limit order. As this increases, the execution probability tends to decrease.
The “mid price divergence rate” is a result of dividing the price difference from the mid price on the above board by the price of the limit order, and is expressed as a percentage. The greater the divergence rate, the lower the execution probability.
“Relative volume at the time of order” means the difference (±) between the volume of the current day when the limit order is generated and the average volume of the past 21 business days at the same time. If the trading volume on the day is relatively large, the subsequent trading volume is expected to increase, so the execution probability is also expected to increase.
“Order time return” means a ratio (±) between the price at the time when the limit order is generated and the previous day's closing price. When the current price is lower than the previous day's closing price, the execution probability of the buy order is increased due to the moment effect, and in the opposite case, the execution probability of the sell order is considered to increase.
“Relative volume for the last 5 minutes” means the difference (±) between the volume for the last 5 minutes at the time the limit order is generated and the average volume for the past 21 business days at the same time.
“Recent 5 minute return” means the ratio (±) of the price when the limit order is generated and the price 5 minutes before.

The order attribute calculation unit 28 calculates the attribute information for each limit order based on the accumulated data in the market data storage unit 14 and stores the result in the order attribute DB 30 (S24).
FIG. 14 shows an example of the registration.

ｐ ：執行確率
β_i：モデルパラメータ
ｘ_i：共変量 Next, the probability model generation unit 32 is activated, and the model parameter (regression coefficient) based on the limit order quantity and the determination result stored in the determination result DB 26 and the attribute information of each limit order stored in the order attribute DB 30 Is calculated (S26).
That is, the execution probability of limit orders is calculated using a binomial logit model (logistic regression model) with 11 covariates as shown in the following equation.

p: Execution probability β _i : Model parameter x _i : Covariate

For this reason, for each limit order stored in the determination result DB 26, the probability model generation unit 32 substitutes 1 for the execution probability p in the case of “confirmed” and 0 in the case of “uncommitted”. By substituting the order quantity stored in the determination result DB 26 for the variable x ₁ and the attribute information stored in the order attribute DB 30 for x _{2 to} x ₁₁ , and performing logistic regression on these multiple sets of samples, A set of parameters {β _i } is determined and stored in the probability model DB 34 (S28).
FIG. 15 shows an example of registration, in which a set of model parameters is registered for each brand and trade type.

The model parameter is used when the execution probability calculation unit 36 calculates the execution probability of a specific limit order.
That is, when the trader transmits an execution probability calculation request specifying the brand, trade type, price, and quantity from the transaction terminal 42 to the execution probability calculation unit 36, the execution probability calculation unit 36 that has received the request (S30) An execution probability model (a set of model parameters β _{0 to} β ₁₁ ) matching the brand and trade type is extracted from the model DB 34 (S32).
Next, the execution probability calculation unit 36 calculates necessary attribute information (covariates x _{2 to} x ₄ , x _{6 to} x ₁₁ ) from the market data storage unit 14 (S34). At this time, the price of the limit order input by the trader is used to calculate the cumulative priority book balance (x ₆ ) and the middle price deviation rate (x ₇ ).
Further, the execution probability calculation unit 36 calculates the remaining time (x ₅ ) by subtracting the current time from the transaction end time (advance or close) on that day (S36).
Next, the execution probability calculating unit 36 calculates the execution probability p by substituting these attribute information (x _{2 to} x ₁₁ ) into the above formula 1 together with the limit order quantity (x ₁ ) input by the trader. (S38).
This calculation result is transmitted to the transaction terminal 42 (S40) and displayed on the display.
As a result, the trader can recognize the execution probability before actually placing a limit order to the market.

In this system 10, as described above, the limit order is virtually extracted from the fragmentary transaction data provided from the stock exchange computer 40, and whether or not each limit order is canceled and whether or not the contract is successful is determined. Judgment and calculation of the execution probability model parameters by statistically analyzing the attribute information related to pending or unfilled limit orders, and based on this, the system calculates the execution probability of the actual limit order .
For this reason, the individual limit orders that are extracted are not necessarily equal to actual transactions, but they comprehensively embody the movements in the market at that time.
In addition, by broadly extracting virtual limit orders from transaction data in this way, it becomes possible to collect a huge number of contracted cases and uncommitted cases, and by deriving model parameters based on these, trading It is possible to construct a probability model that strongly reflects market trends.
As a result, the execution probability can be calculated with high accuracy by inputting the current data and the actual limit order data into the probability model.

  In the above, as the attribute information that affects the execution probability of limit orders, order quantity, best buy price, best sell price spread, spread, remaining time, cumulative priority book balance, midpoint divergence rate, relative volume at the time of order, order Although the example of adopting the point-in-time return, the relative volume for the last 5 minutes, and the return for the last 5 minutes has been described, the present invention is not limited to this, and some combinations of these or other attribute information A probabilistic model can also be generated based on the combination.

It is a block diagram which shows the function structure of the execution probability calculation system of a limit order. It is a schematic diagram which shows the structure of the transaction data transmitted from the computer of a stock exchange. It is a screen layout figure which shows the state which expressed the sales price and the bid price which concern on a specific brand in two dimensions. It is a flowchart which shows the process sequence of the execution probability calculation system of a limit order. It is explanatory drawing which shows the registration information in a market data storage part. It is explanatory drawing which shows the logic at the time of extracting a limit order and cancellation. It is explanatory drawing which shows the registration information in limit order DB. It is explanatory drawing which shows the registration information in cancellation DB. It is explanatory drawing which shows the logic for distinguishing the contract based on a market order, and cancellation. It is explanatory drawing which shows the rule at the time of distributing the share of a contract to each limit order. It is explanatory drawing which shows the rule at the time of distributing the share for cancellation to each limit order. It is explanatory drawing which shows the registration information in determination result DB. It is a block diagram which shows the arithmetic processing performed by the order attribute calculation part. It is explanatory drawing which shows the registration information in order attribute DB. It is explanatory drawing which shows the registration information in probability model DB.

10 Limit order execution probability calculation system
12 Market data acquisition department
14 Market data storage
16 Limit order extractor
18 Limit order DB
20 Cancel extraction part
22 Cancellation DB
24 execution judgment part
26 Judgment result DB
28 Order attribute calculator
30 Order attribute DB
32 Stochastic model generator
34 Stochastic model DB
36 Execution probability calculator
38 Communication network
40 Stock Exchange Computer
42 Trader trading terminal
44 boards
46 Price range

Claims (4)

Order information including at least stock issues, date / time, price, quantity sent from the stock exchange computer and quotation information including at least stock issues, date / time, trade type, price, quantity are sequentially stored in the market data storage unit. Means for storing;
Limit order extraction means for extracting limit order information based on the time series change of the above-mentioned quotation information, and storing at least the name, date, trading type, price, quantity in a limit order storage unit in association with a unique ID;
Cancel extraction means for extracting cancellation information of limit orders based on the time series change of the execution information and the quotation information, and storing at least the brand, date / time, trade type, price, quantity in a cancellation storage unit in association with a unique ID When,
Based on the above execution information and cancellation information, execution determination, cancellation, uncommitted determination for each limit order, each determination result stored in the determination result storage unit,
Means for calculating predetermined attribute information relating to pending and unfilled limit orders based on the determination result and information stored in the market data storage unit;
Means for calculating the remaining trading time based on the date and time of each limit order and the trading end time of the stock exchange;
Statistical processing is performed on the determination result of promise or uncommitted for each limit order, the quantity, the remaining time of the transaction, and the attribute information to generate a probability model for each issue, and stores this in the probability model storage unit Means,
Means for selecting a probability model to be applied based on a stock when a limit order specifying the stock, trade type, price, quantity is entered;
Means for calculating necessary attribute information at the present time of the limit order based on information stored in the market data storage unit;
A means for calculating the remaining time of the transaction based on the current time and the transaction end time of the stock exchange;
A system comprising means for calculating execution probability by applying the limit order brand, trade type, price, quantity, remaining time of transaction, and necessary attribute information to the probability model,
The limit order extraction means is configured such that when the quantity related to the specific brand, trade type, price quotation information stored at the time t stored in the market data storage section increases at a later time t + 1, the brand , Date / time, trade type, price, limit order information for increased quantity,
The cancel extraction means is configured such that the quantity related to the quotation information of a specific brand, trade type and price at a certain time t stored in the market data storage unit decreases at a later time t + 1, and the quantity decreased during this time If the contract information for the minute does not exist, the cancellation information for the relevant brand, date and time, trade type, price, reduced quantity,
The execution determination means determines the success or failure of the execution by assigning a fixed amount in order of priority in terms of time and price for each limit order, and assigns a cancel quantity in order of subordinate in time and price. A limit order execution probability calculation system characterized by determining success or failure of the cancellation, and determining that a limit order that has no determination of execution or cancellation is unconfirmed.   2. The limit order execution probability calculation system according to claim 1, wherein at least one of the attribute information is a cumulative balance of orders that are prioritized in terms of time and price with respect to a target limit order.   3. The bid price according to claim 1, wherein at least one of the attribute information is a bid price divergence ratio obtained by dividing a price difference between a target limit price and a bid price by a limit order price. Order execution probability calculation system. Computer
Order information including at least stock issues, date / time, price, quantity sent from the stock exchange computer and quotation information including at least stock issues, date / time, trade type, price, quantity are sequentially stored in the market data storage unit. Means for storing,
Limit order extraction means for extracting limit order information based on the time series change of the quotation information, and storing at least its brand, date, trading type, price, quantity in a limit order storage unit in association with a unique ID,
Cancel extraction means for extracting cancellation information of limit orders based on the time series change of the execution information and the quotation information, and storing at least the brand, date / time, trade type, price, quantity in a cancellation storage unit in association with a unique ID ,
Based on the above execution information and cancellation information, execution determination means for determining execution, cancellation, uncommitted for each limit order, and storing each determination result in a determination result storage unit,
Means for calculating predetermined attribute information relating to pending and unfilled limit orders based on the determination result and information stored in the market data storage unit;
Means for calculating the remaining trading time based on the date and time of each limit order and the trading end time of the stock exchange;
Statistical processing is performed on the determination result of promise or uncommitted for each limit order, the quantity, the remaining time of the transaction, and the attribute information to generate a probability model for each issue, and stores this in the probability model storage unit means,
Means to select a probability model to be applied based on the stock when a limit order specifying the stock, trade type, price, quantity is entered,
Means for calculating required attribute information at the present time of the limit order based on information stored in the market data storage unit;
Means for calculating the remaining time of the transaction based on the current time and the stock exchange's transaction end time;
A program that functions as a means for calculating execution probability by applying the limit order brand, trade type, price, quantity, transaction remaining time, and necessary attribute information to the probability model,
The limit order extraction means is configured such that when the quantity related to the specific brand, trade type, price quotation information stored at the time t stored in the market data storage section increases at a later time t + 1, the brand , Date / time, trade type, price, limit order information for increased quantity,
The cancel extraction means is configured such that the quantity related to the quotation information of a specific brand, trade type and price at a certain time t stored in the market data storage unit decreases at a later time t + 1, and the quantity decreased during this time If the contract information for the minute does not exist, the cancellation information for the relevant brand, date and time, trade type, price, reduced quantity,
The execution determination means determines the success or failure of the execution by assigning a fixed amount in order of priority in terms of time and price for each limit order, and assigns a cancel quantity in order of subordinate in time and price. A limit order execution probability calculation program characterized by determining success or failure of the cancellation, and determining that a limit order that has no determination of execution or cancellation is determined to be unsettled.

Images